http://wrap.warwick.ac.uk/

Original citation: Liu, Chengwei and de Rond, Mark. (2015) Good night, and good luck : perspectives on luck in management scholarship. The Academy of Management Annals . pp. 1-56 Permanent WRAP url: http://wrap.warwick.ac.uk/75811 Copyright and reuse: The Warwick Research Archive Portal (WRAP) makes this work of researchers of the University of Warwick available open access under the following conditions. Copyright © and all moral rights to the version of the paper presented here belong to the individual author(s) and/or other copyright owners. To the extent reasonable and practicable the material made available in WRAP has been checked for eligibility before being made available. Copies of full items can be used for personal research or study, educational, or not-for-profit purposes without prior permission or charge. Provided that the authors, title and full bibliographic details are credited, a hyperlink and/or URL is given for the original metadata page and the content is not changed in any way. Publisher statement: "This is an Accepted Manuscript of an article published by Taylor & Francis Group in The Academy of Management Annals on 23/11/2015, available online: http://www.tandfonline.com/10.1080/19416520.2016.1120971 .A note on versions: The version presented here may differ from the published version or, version of record, if you wish to cite this item you are advised to consult the publisher’s version. Please see the ‘permanent WRAP url’ above for details on accessing the published version and note that access may require a subscription. For more information, please contact the WRAP Team at: publications@warwick.ac.uk

http://wrap.warwick.ac.uk/http://wrap.warwick.ac.uk/75811http://www.tandfonline.com/10.1080/19416520.2016.1120971mailto:publications@warwick.ac.uk

1

Good Night, and Good Luck:

Perspectives on Luck in Management Scholarship

Chengwei Liu* chengwei.liu@wbs.ac.uk University of Warwick

Mark de Rond

m.derond@jbs.cam.ac.uk University of Cambridge

Abstract

It is not insignificant that seminal contributions to management scholarship have highlighted

luck as an alternative explanation for performance differences between individuals and

organizations. Yet it has rarely taken center-stage in scholarship. The principal purpose of this

paper is to provide a systematic review of the application of luck in the management literature

and in such foundation disciplines as economics, sociology, and psychology. Our analysis

finds five common perspectives on luck: (a) luck as Attribution; (b) luck as Randomness; (c)

luck as Counterfactual; (d) luck as Undeserved; and (e) luck as Serendipity. We outline

various ways in which research on luck may be advanced along each of these perspectives,

and develop an underexplored, sixth, perspective on (f) ‘luck as Leveler’ to provide a possible

solution to such issues as social inequality and (unwarranted) executive compensation.

Keywords: luck, attribution, counterfactual, moral judgment, randomness, serendipity,

chance, management, inequality

* corresponding author

2

Introduction

‘Good night, and good luck’ is how controversial US journalist Edward R. Murrow would

routinely sign off his 1950s television broadcasts. Given the lateness of the hour, the former is

understandable, but why the latter? Why is it that we wish each other good luck? Is it because

however rational we may be, we remain firm in the belief that good intentions and hard work

may not suffice to bring about a desired result? That ours is an unpredictable world?

Despite their best efforts, individuals and organizations can find it difficult to predict the

consequences of their initiatives. When assuming good intentions, how else are we to account

for the corporate duds that regularly headline the popular press? Even with the benefit of

hindsight, those studying organizations can find it difficult to supply robust, comprehensive

explanations for their success. For why is it that some succeed where others fail, or succeed

more often, and for longer, than others do? To this question – arguably the central in

management scholarship – one finds various explanations. The more familiar of these include

exchanges between Mintzberg and Gould on Honda’s success with the Supercub motorcycle

(Mintzberg, Pascale, Rumelt, & Goold, 1996), and on the importance of industry for firm

success between McGahan and Porter (2002) on the one hand, and Rumelt (1991) on the

other. Whereas Rumelt highlighted the predominance of random variations at the macro level,

Mintzberg makes the point that luck had to be an important part of the explanation at the

micro level (for Honda). Both accounts join a steady line of scholarly contributions in the

management literature that explicitly referencing luck as an explanation for performance

differences (Alchian, 1950; Aldrich, 1979; Arthur, 1989; Barney, 1986; Cyert & March,

1963; Denrell, 2004; Denrell, Fang, & Liu, 2015; Hannan & Freeman, 1989; Levinthal, 1991;

Lippman & Rumelt, 1982; Nelson & Winter, 1982; Porter, 1991; Starbuck, 1994). Yet such

references remain the exception rather than the rule: a review of the use of luck in six leading

management journals (Academy of Management Journal, Academy of Management Review,

3

Administrative Science Quarterly, Management Science, Organization Science, and Strategic

Management Journal) suggests that only 2 percent of articles included the word ‘luck’ in the

main text, abstract or title (see Table 1). And the reasons for this may not be hard to find.

After all, how is one to operationalize – let alone draw practical implications from –

something as, well, fickle and haphazard as luck?

Insert Table 1 about here

To not have referenced luck explicitly does not also mean that management researchers

have discounted its importance. Occasionally they have used alternative constructs to

acknowledge something quite similar. Where luck is referenced, its meaning can vary widely.

For some, it is the unexplained variances that lack pragmatic value (Porter, 1991). As Barney

(1997: 17) writes: what prescriptive advice can we give to managers given that the role of

luck is important, ‘that they should “be lucky”?’ For others, luck is essential for explaining

performance differences because randomness in structured environments can produce

systematic patterns (Denrell, 2004; Denrell et al., 2015; Henderson, Raynor, & Ahmed, 2012;

Levinthal, 1991). Still others argue that while good and bad luck can happen to anyone, some

are more prepared than others (de Rond, 2014; Dew, 2009), for example, by being mindful

enough to rebound from bad luck (Weick & Sutcliffe, 2006), or by securing a higher ‘return

on luck’ (Collins & Hansen, 2011). Some even argue that others’ systematic underestimation

of luck can signal profitable opportunities (Denrell & Liu, 2012; Lewis, 2003; Mauboussin,

2012; Powell, Lovallo, & Fox, 2011; Taleb, 2001).

The primary purpose of this paper is to provide a systematic review of the use of luck in

the management literature, and in such related disciplines as economics, sociology and

psychology on which management scholars frequently draw. This review will identify and

4

unpack five distinct interpretations of the role of luck: (a) luck as Attribution; (b) luck as

Randomness; (c) luck as Counterfactual; (d) luck as Undeserved; and (e) luck as Serendipity,

before adding a sixth: (f) luck as Leveler. Each perspective is closely tied to a particular

literature that entails as yet unsolved puzzles that suggest promising directions for future

research. For example, while prior studies in psychology uniformly suggest that people tend

to mistake luck for skill when evaluating performance differences (Baron & Hershey, 1988;

Kahneman, Slovic, & Tversky, 1982; Langer, 1975; Nisbett & Ross, 1980; Rosenzweig,

2007; Ross & Nisbett, 1991), recent studies point at an asymmetry in the evaluations between

successes and failures (Denrell & Liu, 2012; Pizarro, Uhlmann, & Salovey, 2003; Uhlmann,

Pizarro, & Diermeier, 2015). While people appreciate the role of (bad) luck in failures, they

do not necessarily do the same when it comes to explaining success. Explaining this

asymmetry has important implications for organizational learning, executive compensation

and responsibility allocation. Moreover, a common theme among these perspectives is that

people’s systematic underestimation of (good) luck tends to create a positive feedback loop

for only a select few that end up taking all the credit for achievement, for having created

opportunities and accumulating wealth even when undeserving (Bebchuk & Fried, 2009;

Frank & Cook, 1995; Lynn, Podolny, & Tao, 2009). This suggests an underexplored

perspective on luck as Leveler that might help attenuate such problems as exacerbating social

inequality and unwarranted executive compensation. We will elaborate on these, and on other

relevant, examples.

The paper is structured in three sections. First, we examine how luck is typically used in

management scholarship, and in such ‘root’ disciplines as psychology, sociology, and

economics. Based on this review, we outline a potential research agenda for each of the

perspectives on luck we identify. We conclude by examining similarities and differences

5

among these perspectives on luck and elaborate their implications for management research

and also teaching.

What Is Luck?

The New Oxford Dictionary of English (NODE) defines luck as: ‘good or bad things that

happen to you by chance, not because of your own efforts or abilities’. Implied in this

definition are three characteristics. First, luck is a psychological attribution people use to

respond to observed events (Hewstone, 1989; Kelley, 1971; Weiner et al., 1971). That is to

say, people are likely to attribute an event to luck if they consider that the event happened by

chance or randomly (de Rond & Thietart, 2007). Second, attributing an event to luck implies

that the event has a salient evaluative status, which can be either ‘good’ or ‘bad’. Whether

luck is ‘good’ or ‘bad’ is beholden to the observer: it is subjective and context dependent

(Rescher, 1995: 32). Third, whether luck is ‘good’ or ‘bad’, or plays any role at all, often

depends on when one takes stock. As time passes, and with the unfolding of events, one’s

assessment of luck can change dramatically (Rescher, 1995). The same event may be

interpreted differently depending on the information available, the situation in which the

explanation is offered, or the motivation for providing the explanation (Runde & de Rond,

2010). To illustrate, imagine the unpalatable: a car crash with multiple casualties. Let us

assume that an exhaustive subsequent investigation finds the crash to have been a freak

accident, meaning that chance played an important role. In terms of our definition, the

scenario fits the three characteristics. The crash lacked obvious intentional design. Also, it

produced salient outcomes that can be evaluated insofar as several people losing their lives,

and others their relatives. And it is also true that the crash needs not necessarily be attributed

to bad luck. Consider that A is a person who died in the crash, B is a person who would have

involved in the crash if he was not delayed by a call and thus had a narrow escape, and C, a

6

person who was not directly involved in the crash and is the unfaithful wife that A was going

to divorce. A, apparently, can be considered unlucky to have died in this accident. B can be

considered lucky to have avoided the crash, even if B may have initially thought himself

unlucky to have been delayed. C is unfortunate to lose her husband in the crash, but if she is

now able to inherit all her dead husband’s property, which she would not have been entitled to

after divorce, perhaps she (and her lover) might be considered lucky. An observer who had

just been through a painful divorce may even consider A to have been lucky to avoid a

potentially difficult process. Perceptions of luck associated with the same crash can be

interpreted wholly differently and illustrates the third characteristic of luck attribution: luck

attribution is always subjective and can change according to the information available and

with a better understanding of the motivations behind the attribution.

The definition of luck used in this paper is consistent with the NODE and aforementioned

three characteristics of attribution. The purpose of the analysis that follows is to show how the

application of luck in the current management literature can broaden and enrich our

understanding of how the interactions between chance, context and human interventions are

relevant to management. We proceed by outlining five common uses of luck: (a) luck as

Attribution; (b) luck as Randomness; (c) luck as Counterfactual; (d) luck as Undeserved; and

(e) luck as Serendipity, before adding a sixth: luck as (f) Leveler.

Five Perspectives on Luck in the Management Literature

Luck as Attribution

To treat luck as attribution is its most typical application in the management literature.

According to Attribution Theory (Hewstone, 1989; Kelley, 1971; Weiner et al., 1971), people

tend to attribute observed outcomes to four possible factors: skill, effort, task difficulty and

luck. Consistent with our dictionary definition, people are more likely to attribute an observed

7

outcome to luck when the cause of the outcome is considered to be external, unstable, and

uncontrollable.

Attribution biases are the focus of ‘luck as attribution’ in the management literature. For

example, when evaluating own performances, self-serving bias suggests that people tend to

attribute their successes to skill or effort and consider failures to be bad luck (Miller & Ross,

1975). Such attributions can lead to over-learning from successes and under-learning from

failures, resulting in an illusion of control (Langer, 1975) and overconfidence (Camerer &

Lovallo, 1999; Hogarth & Makridakis, 1981). People are also notoriously incompetent in

interpreting outcomes that involve randomness. For example, both the ‘gambler’s fallacy’ and

the ‘hot-hand fallacy’ suggest that people hold illusory beliefs that a random sequence entails

a pattern (Ayton & Fischer, 2004; Gilovich, Vallone, & Tversky, 1985; Tversky &

Kahneman, 1974). Moreover, instead of seeing luck as an external, random element, people

often interpret luck as a personal characteristic (Darke & Freedman, 1997; Maltby, Day, Gill,

Colley, & Wood, 2008), and believing it can be manipulated (Tsang, 2004a, 2004b).

Attribution biases are also prevalent when evaluating others’ performances. People tend to

evaluate a decision based on its realized outcome rather than the quality and situations at the

time the decision was made (Baron & Hershey, 1988). Partly as a result of the Halo Effect

(Rosenzweig, 2007), executives whose decisions resulted in success are treated as heroes and

those who failed as villains even when their decisions are identical (Dillon & Tinsley, 2008).

Such an underestimation of the role of luck is consistent with the Fundamental Attribution

Error, or the tendency of people to over-attribute outcomes to dispositional factors such as

skill than to situational factors such as luck (Gilbert & Malone, 1995; Ross & Nisbett, 1991).

Moreover, selection bias suggests that people tend to under-sample failures and focus on

survivors (Denrell, 2003). This implies that people draw lessons from ‘lucky’ survivors even

when the lesson learned can be detrimental to future performance.

8

To ignore regression to the mean is another common bias when evaluating observed

performances. More extreme performances tend to be followed by less extreme ones because

extreme performance tends to be associated with extreme luck and such luck is unlikely to

persist, suggesting that future performance should regress to the mean (Harrison & March,

1984). Yet people appear unconvinced by statistical accounts of performance change. Instead,

they tend to generate their own causal explanations for such changes. For example, regression

to the mean suggests that poor performance is likely to be followed by improvement whereas

good performance is likely to be followed by a decline. However, we tend to reward others

when they perform well and punish others when they perform poorly. This can lead to

superstitious learning insofar as we may wrongly conclude that being nice to others can cause

decline and being nasty to others can cause improvement (Kahneman, 2011; Kahneman &

Tversky, 1973). Nevertheless, these changes in performances may only reflect regression to

the mean that requires no causal explanation.

Future directions. The aforementioned attribution biases offer some reasons for people’s

misperceptions of luck. People are likely to underestimate the role of luck (e.g., randomness

and situational factors) in observed outcomes. One possible reason for this bias is that people

tend to apply cognitive shortcuts by substituting a difficult question (e.g., what is the

unobserved level of skill of an executive) with an easy one (e.g., what is the observed level of

performance of an executive?) (Kahneman, 2011). Such substitutions can be a useful

heuristic, not only because they often save time and energy, but because they could be correct

(e.g., higher performers are likely to be more skilled) and thus entail Ecological Rationality

(Goldstein & Gigerenzer, 2002). That said, such substitution should be avoided in that errors

can be costly. This suggests at least three questions to guide future research. First, when is it

desirable to apply cognitive shortcuts when evaluating performance differences? The answer

9

to this question will depend on the difference in cost between two possible errors: a false

positive (e.g., mistaking luck for skill) and a false negative (e.g., mistaking skill for luck). As

we have seen, attribution biases suggest that people tend to err towards false positive errors

when evaluating performance. This may entail important motivational functions, e.g., people

maintain esteem and appetite for taking risk when engaging in self-serving biases. By

contrast, a tendency to err on the side of false negatives implies inaction, which is often more

costly than action in evolutionary processes (Richerson & Boyd, 2005). This suggests that we

may be hard-wired to make more false positive errors that, even if imprecise, are often useful.

The difficulty is that cognitive shortcuts that might have once been useful for our ancestors

may now lead to errors that are too costly to bear in modern societies. Think, for example, of

the Columbia Space Shuttle disaster, which resulted from evaluating lucky near-misses as

successes (Dillon & Tinsley, 2008), or of financial crises that resulted partly from rewarding

analysts’ and traders’ luck (Hilary & Menzly, 2006), or even of (dubious) entrepreneurial

ventures which have arisen from the tendency to misevaluate chance as successes (Lowe &

Ziedonis, 2006; Shane, 2008). In sum, mistaking luck for skill may be a useful general human

attribution tendency but can lead to costly errors in more today’s complex situations. Future

efforts can focus on specifying the conditions under which such biases can lead to undesirable

outcomes before suggesting remedies.

Second, how do aspirations interact with attributions about luck?1 Aspirations play an

important role in performance evaluations (Cyert & March, 1963). People adjust aspirations

based on performance feedback and interpret outcomes as successes (or failures) if current

performance is above (or below) such aspiration. Given that people tend to attribute success

to skill and effort, and failures to luck or circumstance, they are also more likely to attribute

outcomes to (bad) luck when entertaining high aspirations, because high expectations are

                                                                                                               1 We thank one of the reviewers for pointing out this interesting direction.

10

more likely to be followed by perceived failures and disappointments. By contrast, low

aspirations are likely followed by perceived successes, and by the attributions that skill and

hard work caused these successes. Such asymmetrical attribution patterns can entail two

processes: (a) people with lower aspirations tend to improve their skill because they believe

skill matters in their successes, whereas people with high aspirations tend not to improve their

skill because they believe skill does not matter in their failures; (b) over time the aspiration

levels will likely converge because the improved skill of the people who might initially have

low aspirations will result in successes that boost these aspirations in the future; the under-

exploited skill of those who might initially have high aspirations will result in more failures

that subsequently decrease their aspirations. In sum, people’s perceived successes and failures

depend on their aspirations, and different aspirations can lead to asymmetrical luck

attributions that have implications for subsequent actions. These processes are just some of

the many possibilities of how aspiration can interact with attribution. Since successes and

failures are rarely absolute but depend instead on aspirations and social comparison (Cyert &

March, 1963; Festinger, 1954), incorporating the role of aspirations in attributions about luck

is likely to be promising in any future research agenda.

Third, how are we to de-bias our misperceptions of luck? It may not always be

necessary to de-bias imprecise attributions that entail useful functions, but it desirable to help

people resist the temptation of substituting a difficult question with an easy one in cases when

the error can be costly. Prior research has suggested remedies for this, for example, by leading

people to believe the stakes are high (Dillon & Tinsley, 2008) or by encouraging an outside

view (based on evidence such as statistics) instead of an inside view (based on intuition and

guts feeling) (Kahneman & Lovallo, 1993). Making people aware that they are subject to

attribution biases is necessary for all of these de-biasing techniques to work, but arriving at an

adequate level of understanding may be difficult in practice. People might come to understand,

11

from first-hand experience, that using intuitive attributions often works well (Hertwig, Barron,

Weber, & Erev, 2004) and extend this technique to less familiar situations, but may find that

applying their newly gained knowledge in different situations leads to a much less positive

outcome (March, 2006). Moreover, the more uncertain the world becomes, the more people

seek and rely on apparently guaranteed solutions (Gimpl & Dakin, 1984). This suggests that

simple de-biasing techniques are unlikely to convince people to deviate from their intuitions.

In accordance with recent research on “nudging” (Thaler & Sunstein, 2008), instead of

“changing mindsets”, it may be more effective to “change context” in an effort to direct

people’s decisions towards better outcomes (Dolan et al., 2012). Nepotism in succession, for

example, has been shown to hurt family business performances (Giambatista, Rowe, & Riaz,

2005). Recent study demonstrates that nepotism may result from ‘bad luck’ having been

augmented in an asymmetrical process of impression formation between family and non-

family members (Liu, Eubanks, & Chater, 2015). This suggests that de-biasing – or changing

mindsets – may not suffice in altering nepotism. Instead, changing contexts by reconfiguring

social ties, for example, is more likely to be effective in addressing this challenge (Liu et al.,

2015). Future research should take into account how techniques of nudging (or changing

decision contexts) can complement conventional approaches of de-biasing (or changing

mindsets) to help people attenuate the undesirable consequences resulted from

(mis)perceptions of luck.

Luck as Randomness

A second use of luck in the management literature is to highlight the random nature of

behaviors in organizations and management (Starbuck, 1994). Even if people have intentions,

and make conscious (or non-random) choices based on these intentions, studies show that

12

outcomes can still appear to be dominated by random processes. Below we discuss four main

sources of randomness in organizations.

First, organizational outcomes appear random partly because outcomes are influenced by

external events that managers have little control over (Pfeffer & Salancik, 1978). Corporate

success is influenced by the activities of competitors (Hannan & Freeman, 1989), the

government (Pfeffer & Salancik, 1978), and by external events such as fluctuations in

exchange rates (Bertrand & Mullainathan, 2001). An innovative firm might be unlucky and

launch a product ‘on budget and on schedule’ only for this to coincide with a recession; a

more favorable timing might have propelled this same product, and firm, to success. As Bill

Gates admitted ‘Our timing in setting up the first software company aimed at personal

computers was essential to our success ... The timing wasn’t entirely luck, but without great

luck it wouldn’t have happened.’ (cited in Mauboussin, 2012, p.13).

A series of seminal studies on sources of variance in corporate profitability (McGahan &

Porter, 2002; Rumelt, 1991) illustrate the importance of events beyond managerial control.

Significantly, they find that as much as half of variations in performance cannot be explained

by firm or industry attributes (McGahan & Porter, 2002). The unexplained proportion of

variance is larger, in most studies, than the proportion of variance explained by any single

factor. In some studies, the unexplained proportion is higher than the sum of the variance

accounted for by all of other factors (McGahan & Porter, 2002). This implies that much of the

variance in profitability cannot be explained by the factors that tend to be the foci in strategy

textbooks.

Second, the outcome of carefully planned behavior would appear to be random if choices

were based on inaccurate forecasts or on an incomplete understanding of means-ends

connections. Empirical studies of forecasting accuracy show that predicting important

business outcomes is challenging (Durand, 2003). The average absolute percentage error, i.e.,

13

(forecast - outcome) / forecast, in forecasts of macro-economic quantities (e.g., inflation,

exchange rates, unemployment) by economists and analysts is about 20% (Armstrong &

Collopy, 1992). Forecasts about demand and product success are even less accurate, with an

average absolute percentage error of close to 50% (Fildes, Goodwin, Lawrence, &

Nikolopoulos, 2009). For fast-moving consumer goods, such as movies and records, even the

best methods have an absolute percentage error of 70% (Lee, Boatwright, & Kamakura,

2003).

Forecast inaccuracy limits how much theories that emphasize persistent firm differences

can explain. If demand changes in ways that are difficult to forecast, profitability will only be

weakly persistent, even if firm capabilities or costs are highly persistent. Forecast inaccuracy

also partly explains why firm growth is nearly random (Geroski, 2005). Capable but unlucky

firms who bet on the wrong product will not grow, while firms with weak capabilities who

happen to bet on the right products will, and this explains why growth rates are almost

random (Coad, 2009).

Third, the outcome of organizational decisions may appear random when events are

decoupled from the intentions of those who are supposed to be in charge, and this will remain

the case even in stable and predictable environments. Managers have less control over

important determinants of competitive advantage, such as culture and capabilities, than

generally thought (Hambrick & Finkelstein, 1987; Pfeffer & Salancik, 1978). Managers may

choose wisely among alternative strategies, but the strategy that is implemented may be very

different from their initial intent (Mintzberg & Waters, 1985).

Realized outcomes may differ from intended ones because those implementing firm

strategy may have different incentives, may not understand what is required, or believe they

know better (Powell & Arregle, 2007). Inertia is another reason why realized outcomes can

differ from those intended. Change programs with the aim of reforming practices and

14

implementing new routines often fail due to inertia (Hannan & Freeman, 1984). Even if

organizations do manage to change, the environment may change faster than organizations

can adapt. The complicated nature of interactions in organizations and markets can also lead

to unanticipated consequences (Perrow, 1984) and changes in one part of the organization

may lead to adjustments in other parts (Hannan, Pólos, & Carroll, 2003). Such indirect effects

may make the impact of any intervention difficult to forecast. Decision processes and conflict

within organizations can also lead to outcomes that are neither understood nor intended

(Aldrich, 1979; Cohen, March, & Olsen, 1972). Finally, people in organizations make

mistakes that can have significant effects. For example, two Harvard economists dramatically

exaggerated the negative impacts of a high debt ratio on GDP growth (Reinhart & Rogoff,

2010). They later acknowledged a mistake with the Excel coding they used which had

”averaged cells in lines 30 to 44 instead of lines 30 to 49” (Herndon, Ash, & Pollin, 2014: 7),

excluding five countries from the analysis. Millions of people’s lives were impacted due to

austerity measures justified by this research.

Fourth, an important but often neglected source of randomness in business is competition.

Competition leads to randomness because it removes obvious opportunities and equalizes

expected returns. Samuelson (1965) provided the first formal demonstration of how

competition between skilled and rational actors in financial markets can lead to randomly

fluctuating asset prices because of equalized returns. In efficient markets, prices reflect all

available information and only change when new information becomes available that could

not have been anticipated based on past data. Stated differently, prices should only react to

unexpected news. The implication is that price changes will be unpredictable and uncorrelated

over time, consistent with empirical evidence. Stock prices are, largely, unpredictable (Fama,

1970). Empirical studies of earnings announcement also show that stock prices mainly react

to unexpected changes in earnings (Beaver, 1968).

15

Competition also leads to randomness due to strategic uncertainty. Even in settings without

uncertainty about external events, uncertainty will exist due to the inherent difficulty of

forecasting the moves of rational competitors (Camerer, Ho, & Chong, 2004). Consider, for

example, the following entry game: a firm can enter three different product markets: A, B, or

C. There is high demand for product A (with 100 consumers), moderate demand for B (30

consumers), and little demand for C (10 consumers). Suppose profitability simply depends on

demand and the number of firms joining a given market: firms in market j will make a profit

equal to demand divided by the number firms in market j. There are 100 firms that

contemplate entry and make their decisions simultaneously. What market should they choose

to join? The decision depends on what the other firms will do, but what they actually choose

to do is likely to depend on their forecasts of what others firms will do.

The only equilibrium in this game is a mixed strategy: the probabilities of joining markets

A, B, and C are 100/140, 30/140 and 10/140 respectively. This equilibrium ensures that

expected profitability is the same in each market and an important implication of such a

mixed-strategy is that profitability will vary randomly between different realizations of the

game, even if all players are rational and adhere to this mixed strategy equilibrium. This

simple game illustrates how competition introduces strategic uncertainty. Even if there is

nothing uncertain in the specification of the game (e.g., no uncertain external events impact

profitability), there is strategic uncertainty that implies that profitability will vary, seemingly

by luck, over time.

Future directions. Randomness is often an endogenous organizational outcome produced by

intentional actors. Since random processes dominate many organizational phenomena, it is

sensible to attribute their associated outcomes to luck. An interesting observation is that this

perspective on luck does not tie to a particular theory in the management literature. Rather,

16

the importance of randomness is highlighted in several different research streams, such as in

decision-making (Cohen et al., 1972), in evolutionary modeling (Nelson & Winter, 1982), in

studies of the distributions of firm growth rates (Geroski, 2005), and in studies of diffusion

(Salganik, Dodds, & Watts, 2006), CEO effects (Fitza, 2014) and competitiveness (Powell,

2003). While these contributions are not directly connected, the recurrent theme of how

randomness in structured environment can produce systematic patterns qualifies a “random

school of thought in management” (for a review, see Denrell et al., 2015). Here we discuss

three directions for future research.

First, selection is an underexplored process that can increase the importance of luck in

outcomes. Selection tends to amplify randomness because it reduces skill differences by

removing weak competitors (Barnett, 2008; Barnett & Hansen, 1996), therefore reducing the

signal-to-noise ratio. March and March (1977) argue that “almost random careers” are an

expected consequence of sorting in organizations that reduces the heterogeneity in skill

among managers, especially at the top. If only sufficiently skilled managers make it to the

next level, the difference in skills among managers who make it to the top will be small.

Unless the variability due to noise (e.g., resulting from external events or unpredictable

subjective performance evaluation) is also reduced, the proportion of variance in performance

due to unsystematic random variation will play an increasingly important role in selection

processes (Thorngate, 1988). Similarly, if selection reduces the variability in firm

productivity (Syverson, 2011), the proportion of variance explained by productivity will

decline. This implies that more extreme performances are associated with greater degrees of

luck, and that tougher selection criteria can lead to less qualified actors being selected. Future

research can also extend this line of research and develop more effective selection

mechanisms that depend less on luck, particularly for high-level executives.

17

Second, management researchers should incorporate randomness into theory building in

order to develop stronger null models when examining hypotheses (Schwab, Abrahamson,

Starbuck, & Fidler, 2011; Starbuck, 1994). Explanations relying on randomness might seem

unfalsifiable: one could always claim that something was due to chance, but how could such a

statement ever be tested? By making parsimonious assumptions that there is no difference

among actors, these ‘naïve models’ usually make more rigorous and detailed predictions than

other theories (Schwab et al., 2011). Many theories in management only make point

predictions about the sign of a coefficient in a regression (e.g., we theorize that the effect of x

on y is positive...). Theories postulating randomness at the micro-level make predictions

about the distribution of outcomes, thus allowing more opportunities for the theory to be

falsified. This also implies that empirical analyses that can reject a naïve model that assumes

no systematic difference among actors or firms can provide more rigorous support for

management theories. Management researchers should take randomness more seriously,

particularly when luck dominates in phenomena managers care about, such as mergers and

acquisition, persistence in performance, and innovation.

Third, given the empirical support for random variation, and the wide range of phenomena

that have been explained by models relying on random variation (Denrell et al., 2015), it may

be possible for randomness to go beyond ‘naïve models’ and acquire a more prominent role in

management theory. In explaining an empirical regularity, a management scholar should

consider explaining this regularity as a result of random variation in a structured system. In

other words, it makes sense to start the search for an explanation by developing a model

relying on random variation. For example, in the area of judgment and decision-making,

Hilbert (2012) explored how a model of randomness (as unbiased noisy estimation) could

explain several judgment regularities that had previously been attributed to cognitive biases

by developing a formal model from which all regularities could be derived. Can a similar

18

unifying formal framework be developed to explain regularities in performance persistence,

career development, firm size, risk taking, and survival? Chance models exist for each of

these regularities (e.g., Denrell, 2004; Denrell, 2008; Denrell & Liu, 2012; Denrell & Shapira,

2009; Geroski, 2005; Levinthal, 1991; March & March, 1977) but these separate models have

not yet been integrated theoretically. Can a formal framework be developed from which all of

these regularities could be derived? Can identical assumptions about probability distributions

be used, and ideally, the same parameter values? A unified chance model to explain various

phenomena central to the field of management would seem a tantalizing prospect, and well

within the realm of possibility.

Luck as Counterfactual

Several management scholars have broadened the application of luck by including

consideration of counterfactuals (Durand & Vaara, 2009; March, Sproull, & Tamuz, 1991).

Thus, an event can be considered to be a matter of luck if it only happens in the realized world

but not in most possible counterfactual worlds (Pritchard, 2005). That is, realized history is

not necessarily efficient (Carroll & Harrison, 1994) and can be considered as drawn from a

pot of possible histories (March et al., 1991). If one could rerun the draw, how likely is it that

an alternative history to that realized could be obtained (Tetlock & Belkin, 1996)? If

counterfactual simulations show that the realized history is, in fact, an unlikely outlier in the

distribution of possible histories, what actually happened can be considered to be luck.

The analysis of counterfactual histories can be problematic because the course of history is

often sensitive to changes in initial conditions and these changes can be augmented in a path

dependent process (Denrell, Fang, & Zhao, 2013; Dierickx & Cool, 1989; Page, 2006). For

example, observed performance differences may seem to result from differences in skill rather

than luck. But if we consider the developmental process of skill, differences in skill may be

19

due to small differences in initial conditions, for example, as being at the right place at the

right time. An exceptional performer may be better than any of her counterparts in realized

history (i.e., what actually happened) but may not have been able to acquire those very skills

in most other, counterfactual histories (i.e., what could just as likely have happened). The

exceptional performer may be better than others but her acquisition of superior skill can be

attributed to luck (Pritchard, 2005; Pritchard & Smith, 2004; Teigen, 2005), or situational

factors (Frank & Cook, 1995).

Consider an example popularized by Malcolm Gladwell: Ice hockey is easily the most

popular professional sport in Canada (Barnsley, Thompson, & Barnsley, 1985). Many

Canadian children aspire to become a professional hockey player, but how can this be

achieved? Research has found a robust empirical regularity in the profile of Canadian

professional hockey players: in every elite group of hockey players studied, at least 40 per

cent were born between January and March (Barnsley et al., 1985). This regularity seems to

suggest that those born between January and March are more talented at playing hockey than

the others and the secret of becoming a professional hockey player in Canada lie in birth dates

(Gladwell, 2008). This example is actually quite a useful illustration of how luck is amplified

by path dependency. High performers from each age group of hockey-playing Canadian

children are selected and groomed for inclusion at the next level. But there is a rule: the cut-

off age for each new hockey league is the 1st of January. This means that those who are born

in the first three months are older and likely to have greater physical maturity than their peers

in the same age class. They are more likely to be chosen to play more often and at higher

levels, where they will have better teammates, better training, and more game experience

(Pierson, Addona, & Yates, 2014). Their advantage is not so much that they are innately

better at hockey, but only that they are older and stronger. Nevertheless, after a few years of

this selection process and the advantages that come from it, the players who are born in the

20

first three months will likely end up being better than their peers who may have had the

potential to have been as good or better.

In the aforementioned example, situational factors such as chance (in this case the birth

date of Canadian children) and context (selection and training in Canadian hockey leagues)

are likely to play more important roles than skill in determining who ends up becoming a

professional hockey player. Both elements of chance and context are beyond the foresight and

control of Canadian children (but not their parents, of course, who have a reasonable

expectation of being able to plan the child’s conception). The initial slight difference in birth

dates, and thus physical maturity, can be augmented in a path-dependent process and produce

huge differences in eventual outcomes. This is occasionally referred to as a ‘relative age

effect’ (Musch & Grondin, 2001). If history could be rerun with slight difference in the initial

condition (e.g., the age cut-off point is 1st of July instead), it is sensible to predict that a large

fraction of the current professional hockey players would have had to settle in different career

paths.

The aforementioned example suggests that luck can have enduring effects in determining

performance differences. The slight advantage gained due to factors beyond one’s control is

usually augmented in a path dependent, rich-get-richer process, i.e., a ‘Matthew Effect2’

(Merton, 1968). Exceptional performances may have little to do with initial levels of skill, but

merely reflect contexts where rich-get-richer dynamics are stronger. Similar processes have

been documented in a variety of research. For example, performance differences can be

considered a matter of luck in the context of wealth accumulation (Levy, 2003; Samuelson,

1989), status hierarchy (Gould, 2002), technology adoption (Arthur, 1989; David, 1985),

cultural markets (Chung & Cox, 1994; Salganik et al., 2006), business competition

(Lieberman & Montgomery, 1988) and even academia (Levitt & Nass, 1989; Merton, 1968).                                                                                                                2 Robert Merton (1968) coined this term from Gospel of Matthew: ‘For to all those who have, more will be given, and they will have an abundance; but from those who have nothing, even what they have will be taken away’. ~Matthew 25:29, New Revised Standard Version.

21

This research all suggests that the eventual performance distribution can reflect an

exaggerated or even distorted initial skill or quality distribution due to luck (Lynn et al.,

2009). Exceptional performers in these contexts should not necessarily impress us because the

winners are likely to have enjoyed early luck of the draw and differences can be seen between

alternative histories.

However, people’s perceptions do not necessarily reflect the role of luck for at least two

reasons. The first arises from the challenges involved in gaining the materials that are

necessary for constructing alternative histories. Perfect counterfactual analysis is impossible if

one cannot specify all of the initial conditions that could have altered the course of history.

This constraint makes counterfactual analysis less practical. The second reason is due to the

way people construct alternative histories in retrospect. Consistent with hindsight bias

(Fischhoff, 1975), the realized history is more salient than others, making people’s

counterfactual imagination anchor in it and underestimate how histories could have unfolded

differently (Byrne, 2005; Kahneman & Miller, 1986; Roese & Olson, 1995). Instead of

mentally simulating possible counterfactual histories, people create positive or affirming

stories that emphasize how human intention and intellect trumps uncertainty and difficulty

(March, 2010). These positive stories offer their tellers and audiences a sense of identity and

practical lessons for future actions, despite the fact that they may not provide the best

reflection of what might have been: ‘a good story is often less probable than a less

satisfactory one.’ (Kahneman & Tversky, 1982: 98). These human-centric stories ‘can be seen

as possibly reflecting elements of human conceit about the role of human intention and

intellect in human behaviors’ (March, 2010: 41). As a result, people often overestimate the

role of skill and underestimate the role of luck in their counterfactual imaginations, mistaking

luck for skill.

22

Future directions. To consider luck in terms of counterfactuals broadens its application by

considering how likely the realized history could have happened differently. By simulating

the distribution of possible histories, a realized history may be attributed to luck if most

alternative histories could have unfolded in very different ways. While offering a useful,

normative approach of conceptualizing luck, rigorous counterfactual analysis is often difficult

because one is unlikely to exhaust all initial conditions that could change the course of

history. On the other hand, the way people construct counterfactual histories is often biased.

We tend to focus on realized history, and do not consider how things might have unfolded

differently, and even when we do, the focus is very much on how changes in human

interventions rather than situational factors could have undone the outcomes. Although these

mental simulations can entail useful function of maintaining motivations and identity, they do

not necessarily accurately reflect the reality. This suggests at least three directions for future

research.

First, future research could examine the role of luck by enhancing the effectiveness of

counterfactual analysis. By computationally or mentally simulating the distribution of

possible histories, one may be able to define the degree of luck of a realized event. For

example, if a risky alternative would lead to bankruptcy in most imagined counterfactual

scenarios, the fact that it is actually realized as a lucky success should not entail attention and

rewards. But such a counterfactual analysis is not easy because controlling for all initial

conditions and their interactions with path dependency is difficult. Recent studies have

suggested novel approaches to address this challenge (Cornelissen & Durand, 2012; Durand

& Vaara, 2009; Vergne & Durand, 2009). For example, some have suggested a ‘contrast

explanation’ approach: one should start by holding all causal factors constant except the one

of interest when developing alternative histories for an event (Tsang & Ellsaesser, 2011).

Another approach suggests that one should start by relaxing key assumptions of existing

23

explanations and develop alternative histories in a more open-ended fashion (Alvesson &

Sandberg, 2011), in contrast to the lab experimental fashion of a ‘contrast explanation’

approach. These approaches all help to generate plausible counterfactual histories more

systematically. Management researchers could apply them when analyzing management

phenomena (MacKay, 2007), as has been done in the such fields as political science (Tetlock

& Belkin, 1996) and military history (Cowley, 2002, 2003), in order to refine these

approaches and enhance the understanding of the role of luck played in management.

Second, failing to mentally simulate accurate alternative histories can be costly. For

example, a shared feature of many organizational disasters is the high number of near-misses,

successful outcomes in which chance plays a role in averting failures, prior to actual disasters

(Perrow, 2011; Starbuck & Farjoun, 2005; Vaughan, 1997). These lucky outcomes are usually

interpreted as successes and people do not consider how the same managerial decision could

have led to a disaster, boosting a false sense of security and appetite for risk taking (Dillon,

Tinsley, Madsen, & Rogers, Forthcoming; Madsen & Desai, 2010; Tinsley, Dillon, & Cronin,

2012). Using an example from Dillon and Tinsley (2008: 1437):

‘On many shuttle missions before the Columbia disaster, foam debris detached from

the shuttle, but luckily never hit a highly sensitive portion of the orbiter. Lacking an

obvious failure, NASA managers interpreted the many near-misses as successes and

accepted the detachment of foam as an ordinary occurrence. What was originally a

cause for concern no longer raised alarms; deviance became normalized.'

Chance averted failure in the cases of these near-misses, but did not with the eventual

disaster. Nevertheless, NASA managers interpreted near-misses as successes and did not

consider how near-misses might easily have turned into disasters. Their perceived risk toward

the foam-related problem was therefore lowered even if the statistical risk of the problem

remained the same. More generally, near-misses signal the vulnerability of the underlying

24

systems and offer opportunities for organizations to fix the problem, but people’s biased

responses mean systems remain vulnerable and the potential for ‘normal accidents’ will

remain (Perrow, 1984). In sum, the way people mentally simulate alternative histories is

problematic, particular for near-misses in interdependent systems. This suggests an urgent

need to examine how to help people to construct less biased alternative histories in order to

improve their ability to evaluate risk. Future research can examine whether encouraging

people to take an outside view (Kahneman & Lovallo, 1993) or inducing people to consider

what might have been could help people evaluate near-misses and chance outcomes in general

more effectively (Liu, 2015).

Third, this perspective on luck offers a new angle on an old debate about skill versus luck.

Earlier studies have shown that rich-get-richer dynamics, and chance elements, make

performance unpredictable and lead to a weak association between ability and success,

implying that success is only a weak signal of skill. But prior studies have not challenged the

idea that top performers are the most skilled and worthy of reward and imitation. More recent

studies show that the belief that the top performers are the most capable is flawed because

exceptional success usually occurs in exceptional circumstance (Denrell & Fang, 2010;

Denrell et al., 2013; Denrell & Liu, 2012). Top performers have been shown to be lucky for

having benefitted from rich-get-richer dynamics that boosted their initial fortune. This implies

that if history were to be rerun, fortune is likely to befall others. Imitating the most successful

in realized history can lead to disappointment or even disasters. Even if one could imitate

everything that the most successful did, one would not be able to replicate their initial fortune

and path dependency. In contrast, less extreme performances may be a more reliable indicator

of skill. These ‘second best performers’ are likely to achieve similar levels of high, but not the

highest, performances in most possible counterfactual histories. There is no rule for becoming

the richest above a certain performance level because achieving exceptional performance

25

usually requires doing something different or new and there can be no recipe for such

innovation (Levy, 2003). The implication is that the more extreme a performance is, the less

one can learn from it because this realized outlier is more likely to indicate unreliability and

could likely have happened differently in alternative histories. This also implies that many top

performers should be dismissed and less extreme performers, i.e., the second best should be

rewarded and imitated.

Luck as Undeserved

A fourth application of luck in the management literature centers around the praise and blame

associated with the unintended consequences of managerial action3. Realized outcomes are

not determined solely by intentional design – uncontrollable and unpredictable factors can

interfere and produce consequences that are decoupled from intention. This suggests that

good intentions do not necessarily lead to good outcomes and vice versa. Nevertheless,

laypersons tend to determine praise and blame primarily based on realized outcomes. This

thus creates a mismatch: well-intended actions or competent managers are blamed for the

failures outside of their control, while ill-intended actions or incompetent managers are

rewarded for achievements that were outside of their control. Most of what falls within this

category sees luck as the residue of intentional design and focuses on how people and

organizations over- or under- reward/punish the actors for their good or bad luck. In particular,

there are three lines of distinct literature that elaborates how managers can receive undeserved

blames or rewards.

The first line of literature relates to executive compensation, and specifically how it is that

executives can receive compensation well beyond what they deserve. This is a central

research topic in the literature of agency theory (Fama, 1980; Jensen & Meckling, 1976)

                                                                                                               3 We thank one of the reviewers for the suggestion of separating this perspective on luck from others.

26

because inappropriate incentive structure can distort the motivation of the executives

(Milgrom, 1981). Two common observations are that (a) executives are often paid for good

luck due to factors beyond their control, such as unanticipated foreign exchange (Bertrand &

Mullainathan, 2001), and that (b) executives tend to be over-paid – more than their current

performances can justify (Bebchuk & Fried, 2009; DiPrete, Eirich, & Pittinsky, 2010; Wade,

O'Reilly, & Pollock, 2006). Research has suggested several reasons for the exaggerated

executive compensation. For example, top executives gain power over the board so the board

conforms to the executives’ self-serving request of high pay (Bebchuk & Fried, 2009). Also,

executives are aware that they can be made scapegoats when performance declines even when

such decline is beyond their control (Boeker, 1992; Wiersema & Bantel, 2006). Requesting a

high compensation package is as if buying insurance for situations when they can be blamed

for bad luck. Overall, this research suggests that the extent of over-pay increases with

executive levels and many top executives’ performances cannot justify the high compensation

they receive. This can hurt the performances of the firm and create concerns about trust and

fairness, which is an important issue we will get back to in subsequent section, when

introducing an underexplored perspective on luck as Leveler.

The second line of literature on ‘luck as undeserved’ originates from philosophy, and

specifically ‘moral luck’ in moral philosophy (Nagel, 1979; Pritchard, 2006; Williamson,

1981). Moral judgments should never depend on luck. After all, it is reasonable to expect

people to take full responsibility for the consequences of voluntary actions and not be blamed

for actions and outcomes that are beyond a person’s control. Yet, moral judgments often

appear to be influenced by luck and circumstances (Nagel, 1979).

Consider two managers facing an identical business scenario – a go/no-go case that

requires a judgment call from them. They both feel that they have conducted enough analyses

and are abiding all the rules, and so both managers decide to go ahead with the plan. In one

27

case, purely by chance, this business plan created an unanticipated, negative side effect that

killed several people while everything went well as planned in the other case4. Normatively,

one should not blame one manager more than the other because the difference in the

consequences, however dramatic, is due to chance. But people tend to judge otherwise – the

former manager tends to be blamed and held responsible for the casualties even when they are

due to bad luck.

Why do we then blame people for their bad luck? Researchers in moral psychology have

proposed several explanations, all of which center around the second-order inferences one

might draw from unlucky incidences. For example, the unlucky incidence that befell the first

manager may indicate false beliefs held by that manager (Young, Nichols, & Saxe, 2010).

Both managers ‘felt’ they had done enough analyses to justify their decision. But perhaps the

bad outcome in the first case indicates that the analyses done by the first manager was simply

not rigorous enough. The first manager may have omitted some important information and

wrongly evaluated the safety procedure of the plan. So the bad luck the manager experienced

would not have been entirely beyond the manager’s control. Another possible inference

involves a judgment of the character of the manager based on observed outcomes (Pizarro et

al., 2003; Uhlmann et al., 2015). The motivation of moral judgment is ultimately about

judging others’ moral character – one should avoid interacting with a person who is perceived

to be immoral. The bad luck the first manager experienced may be informative about

character: being reckless or imprudent, for example, means the manager deserves to be

blamed (Uhlmann et al., 2015). This is also consistent with the research on the belief in luck –

lucky or unlucky outcomes are associated with perceived personal characteristics (Day &

                                                                                                               4 This example is adapted from a now classic example Nagel (1979) developed: two law-abiding drivers, one of whom experienced an unexpected and uncontrollable event-a young child running in front of the car and this driver hit and killed the girl whereas the other driver arrived the destination uneventfully. Philosophers ask: should we blame the former driver more than the latter one? Our adapted example happens in typical management context but the essence is identical to the one developed by Nagel (1979).

28

Maltby, 2003; Maltby et al., 2008). Overall, research on moral luck suggests that people’s

moral judgment is not neutral to luck: laypersons as naïve moralists tie their judgment to the

consequences and discount the role of luck and circumstances. Nevertheless, there is an

ongoing debate about the underlying mechanisms for such regularities in moral judgment.

The third line of literature under this perspective on luck is about how it is people evaluate

executive accountability after extreme failures. Executives and operators are often blamed for

large scale failures (Perrow, 1984) – be they oil spills, nuclear plant disasters, or financial

crises – and expected to assume responsibility and either resign or get fired. Prior studies on

disaster dynamics have emphasized the role of bad luck in failures: system failures can result

from exogenous factors hitting a fragile system. Systems may be so tightly coupled that even

small external shock can cascade and collapse the system due to its interactive complexity

(Perrow, 1984). Consider again how the Columbia Space Shuttle broke apart due to a piece of

foam insulation happened to strike its left wing rather than the other places in prior missions

(Starbuck & Farjoun, 2005). System failures can also result from endogenous processes. Non-

novel interruptions can cumulate to such extent that they create additional interruptions faster

than the executives in charge can fix existing ones (Rudolph & Repenning, 2002), leading to

inevitable system collapse after passing a tipping point. For example, consider how the

Tenerife Air Disaster, the deadliest accident in aviation history, unfolded with several factors

such as terrorist attack, weather, airport capacity together exacerbating the situation to such

extent that the disaster was set in motion and beyond the pilots’ control. Overall these studies

suggest that the situational factors – such as the system characteristics and the external

circumstances – are necessary for producing extreme failures. The implication is that failed

executives do not deserve to take all the blame – they may simply be at the wrong time and

the wrong place.

29

Future directions. This perspective on luck emphasizes a mismatch: realized outcomes do not

necessarily reflect intentions. Nevertheless, people tend to judge the quality of the actor by

outcomes. This suggests that some individuals can be over-blamed for bad luck while others

are over-rewarded for good luck. In particular, top executives can to be over-compensated for

high performances beyond their control and over-blamed for disasters that are also beyond

their control. This creates problems not only for incentive structure, but also for system

robustness: unlucky executives are fired and the system remain fragile, awaiting the next

‘normal accident’ (Perrow, 1984). Moreover, management researchers also recognize the

importance of moral luck in business (Michaelson, 2008; Velamuri & Dew, 2010). One

challenge is that most empirical research on moral luck is based on people’s judgment on

simplified scenarios like the above illustrative example: two similar individual actions lead to

very different outcomes and how people judge the individuals differently based on small

modifications of the scenarios. It is not clear whether these findings are readily applicable for

managerial contexts where uncontrollable forces are often a combination of social influences

and incentive structure. Consider, for example, a case discussed by Arendt (1963), of a

German soldier who served in Auschwitz. He was following orders to kill and torture

prisoners as did most of his peers, and the cost of disobeying the orders might well have been

his own death. To what extent should we blame him for his involvement in the Holocaust? Or

consider a less extreme example: if a manager goes ahead with a mission partly because of an

apporaching deadline and the cost of missing such deadline is likely his job (Perrow, 1984;

Starbuck & Farjoun, 2005), to what extent should he be blamed if the plan happens to go

wrong? These are the sorts of difficult questions about moral luck that have not been

addressed in management. But there are three further directions for future research.

First, when should we blame executives for failure? Normal Accident Theory (Perrow,

1984) suggests that failed executives tend to be over-blamed. However, alternative theories

30

have suggested that the persons in charge should be blamed for extreme failures. Extreme

failures can result from cascading errors that are beyond the control of the executives. But

whether a cascade of errors is triggered is not entirely independent to the executives’ skill.

Clearly, poor skilled executives can exacerbate the situations and make system collapses more

likely. Tightly coupled systems can still perform reliably and be failure-free if the executives

in charge are mindful enough to overcome inertia and be resilient against the unexpected

before additional damages are created (Weick & Sutcliffe, 2006). For example, the Columbia

Space Shuttle Disaster could have been averted had executives at NASA paid attention to,

rather than having normalized, the deviances (e.g., the broken foam insulation) and

interpreted prior near-misses not as successes but a cause for concern (Dillon & Tinsley, 2008;

Vaughan, 1997). Overall these studies on High Reliability Organizations suggest an important

dispositional factor – mindfulness (Langer, 1989) – is crucial for complex organizations with

high-risk technologies to continue performing at a highly reliable fashion (Weick, Sutcliffe, &

Obstfeld, 1999). The implication is that failures are informative about skill - the failed

executives are likely to be the mindless ones who easily succumb to inertia and enable

cascading errors to happen in the first place.

Taken together, this literature adds an important consideration when evaluating the extent

to which failed executives deserve to be blamed: the interactions between the scope for skill

and luck (Denrell, Liu, & Maslach). If cumulated failures do not diminish the scope for skill,

this suggests that higher skilled executives are more likely to exercise their skill and stop

failures from escalating. In this case, a failure is informative about low skill and the failed

executives deserve to be blamed. In contrast, if the exacerbated situations can diminish the

scope for skill because cascading errors overwhelm managerial interventions, eventual failure

is not informative about skill. Rather, it indicates the system is likely to be tightly coupled and

sensitive to external shocks and small errors. Failed executives cannot be blamed here

31

because, however skilled they might be, they could not have averted failure. In particular, the

magnitude of the extreme failures should not be proportional to the blame the failed

executives receive when small errors can cascade and override skill – the magnitude indicates

more about the system characteristics. The interaction between skill (e.g., mindfulness and

resilience) and luck (e.g., exacerbating situations like cascading errors) goes beyond the

simple scenarios discussed in the moral luck literature and should attract more attention from

management researchers.

Second, recent studies suggest an interesting asymmetry in both moral judgment and

performance evaluations. In moral judgment, people tend to discount blame to bad outcome

but not praise to good outcome (Pizarro et al., 2003). Similarly, people tend to appreciate the

role of bad luck for extreme failures but not the role of good luck in exceptional performances

(Denrell & Liu, 2012). A second order inference process may account for this asymmetry:

most people assume that others do (not) want to experience good (bad) luck, or put

themselves in a situation that would increase (decrease) exposure to good (bad) luck. This

suggests that they discount the blame associated with failure because it is against that person's

‘meta-desire’. By contrast, people do not discount the praise associated with success because

that is consistent with that person’s ‘meta-desire’ (Pizarro et al., 2003). While this account is

theoretically sound, more empirical and experimental work is needed to examine the reasons

for such asymmetry, which has potentially important implications for incentive structure and

executive accountability.

Third, when should we reward success? It is sensible to reward higher-achieving

performers when performance is a reliable indicator of skill and effort (Fama, 1980; Milgrom,

1981). Otherwise higher-achieving performers are rewarded for their good luck because their

performances depend more on situational rather than dispositional factors. Nevertheless,

rewarding luck is a common business practice, particularly for the corporate elite (Bertrand &

32

Mullainathan, 2001). On one hand, the difference in skill among the corporate elites is likely

to be tiny (March & March, 1977), implying that their successes and failures are more likely

to result from situational enablers or constraints. On the other hand, people chase a

romanticized perspective on these stars (Khurana, 2002; Meindl, Ehrlich, & Dukerich, 1985),

even when star performances are highly likely to result from uncontrollable factors (Bertrand

& Mullainathan, 2001; Strang & Macy, 2001). In fact, the supposed top performers can be

expected to be less skilled than their lower performing counterparts when extreme

performances indicate unreliability such as excessive risk-taking or cheating (Denrell & Liu,

2012). This implies a large discrepancy between people’s romanticized perspective on how

corporate elites are responsible for firms’ destiny and the reality of how luck dominates the

performances beyond a certain level.

Moreover, the exaggerated high compensation of top executives creates problems for

redistributive justice and endangers the stability of societies. In particular, it hurts the belief in

a just world. Belief in a just world is closely related to perceptions of the extent to which

outcomes should be attributed to luck (Alesina, Glaeser, & Sacerdote, 2001). It is often an

illusion to believe that the world is just. To work hard, for example, does not guarantee the

payoff one deserves when outcomes are largely determined by social connections and

inherited wealth (Piketty, 2014). Such an illusion often entails desirable outcomes because of

a self-fulfilling prophecy: those who believe luck matters less will be motivated to work

harder (Benabou & Tirole, 2006; Gromet, Hartson, & Sherman, 2015). Research shows that

belief in a just world is a prevalent illusion (with notable differences between countries) and

can explain why GDP growth is higher in the United States (which underestimates luck more)

than in many European countries (which underestimate luck less) (Alesina et al., 2001).

However, such misperceptions can backfire: Alesina et al. (2001) also show that

underestimating luck leads to less social spending in the United States. If the majority

33

believes that luck matters less and one should be responsible for one’s own fate, social

spending like expenses on medical care is likely to be lower, potentially decreasing social

mobility and strengthening social inequality. This in turn can lead to social instability because

beyond a certain threshold of social inequality more people will come to realize that the world

is less just than they believed. The exaggerated pay to the executives and bankers, particularly

after the financial crises strengthen this impression as the demonstrations such as ‘Occupy

Wall Street’ would appear to suggest. There may be some rational accounts for why high

executive compensation is useful, but executives should also consider the consequences of

their high compensation packages for society, lest they should decide to react to such felt

injustice in more radical ways. Overall, this discussion suggests that rewards for higher

ranked corporate executives should be proportionally less because luck plays a more

important role in performance at higher levels in corporations. Otherwise the high

compensation of top executives is not only unjustifiable but hurts the belief in a just world

and increases instability in societies.

Luck as Serendipity

A fifth usage of luck in the management literature emphasizes why some people and firms are

luckier than others, or how ‘chance favors the prepared mind’ as Louis Pasteur put it. Good

and bad luck befalls to all, but only some can maximize the return on luck. The focus is not

on chance or luck per se, but on individual or firms’ traits that make them to able to see what

others do not see, a form of “serendipity”. Before we elaborate the role of serendipity in

management context, we introduce its origin, which is a closer correlate of creativity, and

often associated with scientific breakthroughs and such lucky industrial discoveries as Velcro,

X-rays, aspirin, Post-It Notes, the HP Inkjet printer and Scotchguard.

34

The origin of serendipity. Serendipity has its etymological origins in a sixteenth century tale

told in a letter, sent by Horace Walpole to a distant cousin on 28 January 1754. In it, three

princes of Serendip (present-day Sri Lanka), sent by their father to fend for themselves so as

to gain practical knowledge of the world, happened upon a camel owner, distraught for having

lost his precious asset. He inquired as to the whereabouts of his camel. While the princes had

not seen the camel, they were able to render an accurate description of it: it was blind in one

eye, lacking a tooth and lame, was carrying butter on one side and honey on the other, and

was being ridden by a pregnant woman. Their description was accurate enough for the camel

owner to become suspicious. He took them captive and delivered them to the emperor. Upon

interrogation it became clear that the description of the camel had been deduced from

observation alone. They explained that they thought the camel blind in the right eye because

the grass had been cropped only on the left side of the road. They inferred that it was missing

a tooth from the bits of chewed grass scattered across the road. Its footprints seemed to

suggest that the animal was lame and dragging one foot. Also, finding ants on one side of the

road and flies on the other, they concluded that the camel must have been carrying butter on

the ant’s side, and honey on other. Finally, as for the presence of a pregnant woman, a

combination of carnal desires on the part of the princes, and imprints of hands on the ground

sufficed to bring about this final conclusion (Merton & Barber, 2004).

Walpole’s tale is instructive because the princes relied on creativity in recombining

events (that came about by chance or happenstance) and in exercising practical judgment to

deduce ‘correct pairs’ of events so as to generate a surprisingly effective (and, as it happens,

entirely accurate) plot (de Rond, 2014). In contrast to other perspectives on luck, serendipity

points towards a distinct capability, namely that of recombining any number of observations

so as to deduce ‘matching pairs’, or sets of observations, that appear to be meaningfully

related (de Rond & Morley, 2010). Serendipity is the ‘prepared mind’ in Louis Pasteur’s oft-

35

cited quip, and the question one of how organizations, and the individuals inside them, can

increase the likelihood of serendipity occurring. This may be what Porter (1991: 110) had in

mind when suggesting that ‘there are often reasons why firms are “lucky”’.

A closer examination of serendipity also suggests a typology (de Rond, 2014). A first

distinction is that between ‘true’ and ‘pseudo’ serendipity, where one seeks A but finds B, and

where B is ultimately the more highly valued. When it came to the Nobel prize winning

discoveries of PCR and DNA, Mullis and Francis and Crick respectively found what they

were looking for but by way of chance. In each case, the objective remained unchanged, but

the route toward achieving this objective proved unusual and surprising.

Crick and Watson’s discovery of the ‘double helix’ structure of DNA was marked by

various unplanned events such as Watson’s loosely related work on TMV (corroborating

their suspicions of a helical structure), and exchanges with Griffith and Donohue (in

directing them toward the specific, but unorthodox, pairing of bases). Yet they always

knew that they were after the structure of DNA, believing it to contain the secret of life.

Thus, DNA illustrates pseudo-serendipity, insofar as chance events enabled the

unraveling of the molecule, yet these events never caused them to deviate from this

original target (de Rond, 2014: 10).

In pseudo serendipity, A is sought and A is found, but via a route quite different from that

originally envisioned (Roberts, 1989). Thus, in the discoveries of sildenafil citrate (the key

ingredient in Viagra) and penicillin, scientists discovered something else than what they had

been looking for (cf. de Rond & Thietart, 2007).

Both Fleming and Pfizer’s scientists applied creativity and practical judgment in

matching observations of unforeseen events with findings reported by others, and in

selecting which of these combinations might be fruitful. They rightly interpreted

coincidences as meaningful in the context of the knowledge available to them at the time.

36

However, the particle from the mycology labs wafting through Alexander Fleming’s open

window to contaminate a bacterial culture is a random variation, as were the unusual

changes in temperature. By contrast, the unanticipated side effects of sildenafil citrate

surfaced in part as a result of research design; after all, toxicity trials tend to use men

between the ages of 18 and 30, as did Pfizer’s clinical trials (de Rond, 2014: 10).

One can make a further distinction between chance as the unintended consequence of

research design, and chance as pure random variation (de Rond, 2014). Thus, in discovering

sildenafil citrate and PCR, opportunities arose as a direct consequence of the way the study

had been designed: the unintended side effects of sildenafil citrate became apparent precisely

because Phase 1 clinical trials use healthy male volunteers. Likewise, Mullis’ discovery of

PCR relied entirely on his recombination of existing technologies (de Rond, 2014: 10). By

contrast, penicillin and DNA benefited from random chance occurrences: the spore in

Fleming’s dish had most likely wafted in from the mycology labs located one floor down.

And Crick was fortunate to share his office with a crystallographer, who pointed out the flaws

in his original, ‘textbook-correct’, model (de Rond, 2014: 10).

Serendipity in the context of management. In the context of management, we can

conceptualize how chance interacts with individual capability so as to co-produce serendipity

in four different ways (Austin, 1978). The first of these refers to luck that cannot be attributed

to the beneficiary in any meaningful way. This application is a base-line case of luck as

serendipity and is consistent with the application of luck by Jay Barney (1986). A firm’s

performance is determined by the values created by the strategic factors the firm owns.

Superior performance is likely to be founded on the firm’s superior foresight about the value

of its strategic factors and from acquiring these factors for less than they are worth. Otherwise

there should be no abnormal returns - any superior performance should be attributed to good

37

luck because profitability is ultimately traced to unexpected price changes. If strategic factors

are priced correctly, based on all of the available information, price changes only occur when

new, unexpected information becomes available (Fama, 1970). Therefore, this is in fact the

base-line case without serendipity – gains should be attributed to pure luck or windfall that is

independent of the person.

The second variation of luck as serendipity is about how luck favors those in motion –

people who are willing to venture new ways to make progress. While good luck may befall

the lazy and inert, favorable outcomes are more likely to be the result of hard work joined by

chance events, or ‘practice and you get luckier’ (Burgelman, 2003). As Charles Kettering,

former head of research at General Motors, put it: ‘keep on going and the chances are you will

stumble on something, perhaps when you are least expecting it. I have never heard of anyone

stumbling on something sitting there’ (as quoted in Austin, 1978: 15). Moreover, the chance

of success may be low and unforeseeable. But continuing to try will entail the exclusion of

alternatives that would have led to failure, enhancing the chance of success over time. The

implication is that experimentation matters and benefits can arise from exposing oneself to

situations that increase the chance of realizing favorable outcomes.

The third variation of luck as serendipity is about how luck favors those who look inward.

Sometime people or firms can pre-adapt: happen to be endowed with strategic factors that can

be recombined into valuable, idiosyncratic strategic advantage (Cattani, 2005, 2006). The

holdings of these valuable strategic factors often result from unintended consequences. For

example, Cattani (2005) uses the case of Corning to illustrate how an unanticipated use of

fiber optics technology enabled Corning to become one of the leaders in long-distance

communications. Alternatively, some idiosyncratic strategic factors can result from people

opportunistic behaviors when dealing with ambiguity and resource uncertainty (Miner, 1987).

This implies that sustainable superior performances are more likely to result from an accurate

38

understanding of firm-specific resources, even when these resources were acquired by

accident or opportunism (Cohen & Levinthal, 1990, 1994; Makadok & Barney, 2001; Powell,

Lovallo, & Caringal, 2006). By extending these unique resources, firms are more likely to

gain competitive advantages that cannot be easily imitated by competitors (Wernerfelt, 1984).

Consider the case of Bill Gates, the founder of the software giant Microsoft (as quote